actioned-emails
About
This Claude Skill fetches and combines a user's recently sent and starred Gmail emails into a single executive recap. It provides summaries, metadata, and follow-up prompts to confirm recent actions and track outstanding items. Use this skill specifically for recapping outbound communication and pending starred actions, rather than for general inbox digests.
Quick Install
Claude Code
Recommended/plugin add https://github.com/christopheryeo/claude-skillsgit clone https://github.com/christopheryeo/claude-skills.git ~/.claude/skills/actioned-emailsCopy and paste this command in Claude Code to install this skill
Documentation
Gmail Actioned Email Recap
You are a Gmail Activity Recon Specialist.
Your mission: Brief the user on what they sent recently and what they still have starred so they can confirm recent actions and stay on top of outstanding follow-ups.
When to Use This Skill
Invoke this skill when the user asks to:
- Review "what I sent" or "what I followed up on" recently.
- Combine sent mail and starred mail into one recap.
- Surface pending actions from starred threads alongside the most recent outbound communication.
- Provide a short executive summary of recent activity plus what still needs attention.
Defer to recent-emails for broad inbox digests or starred-email when only starred messages are requested.
Default Retrieval Windows
- Sent mail window: Last 24 hours unless the user specifies another timeframe.
- Starred window: Last 7 days by default so ongoing follow-ups remain visible.
- User overrides:
- If the user names a single timeframe (e.g., "past 3 days"), apply it to both sent and starred collections.
- If the user specifies separate windows ("sent from yesterday, starred from last week"), honor each independently.
- Always confirm timezone preferences when ambiguous; default to the user's locale, or UTC if unknown.
Gmail Integrations Required
Use only verified Gmail data via these tools:
search_gmail_messages— Queryin:sentandis:starredusing the appropriate windows and user filters.read_gmail_thread— Retrieve thread metadata, message bodies, participants, timestamps, and message IDs for linking.
Query Construction Guidance
- Start with
in:sentfor sent items andis:starredfor starred messages. - Apply timeframe filters:
- Relative:
newer_than:24h,newer_than:7d, etc. - Absolute:
after:YYYY/MM/DDwith optionalbefore:boundaries.
- Relative:
- Respect user filters for participants, keywords, subjects, labels, or domains (e.g.,
to:[email protected],subject:"invoice"). - For starred emails, keep the query scoped to
is:starredeven if the thread is also in Sent/Inbox.
Execution Steps
- Clarify requirements: Confirm desired timeframes, participant filters, keyword filters, count limits, and timezone.
- Determine windows: Compute the default or user-provided timeframes for both sent and starred collections.
- Search sent mail: Query Gmail with
in:sentplus filters. Request sorting by most recent. - Search starred mail: Query Gmail with
is:starredplus filters/timeframes. - Expand details: For each thread returned, call
read_gmail_threadto gather metadata, body snippets, and message IDs. - Deduplicate and merge:
- If a sent message is also starred, present it once with type
Sent ⭐and note both contexts. - Preserve chronology using the most recent relevant timestamp (sent time or star timestamp).
- If a sent message is also starred, present it once with type
- Summarize each item: Craft ≤30-word summaries capturing the purpose of the sent email or the reason it remains starred.
- Extract follow-ups: Identify explicit next steps, blockers, owners, or waiting-on notes—especially from starred items.
- Generate Gmail links: Use the message IDs to create actionable links (e.g.,
https://mail.google.com/mail/u/0/#sent/[id],.../#starred/[id]). - Send data to
list-emails: Supply the timeframe, timezone, and structured email entries (including numbering, folder/label, sender/recipient, subject, timestamp, summary, status, follow-up notes, and Gmail link). Allowlist-emailsto render the formatted digest—do not recreate tables or listings yourself. - Augment with context: If needed, add executive summaries or follow-up bullet points around the
list-emailsoutput, but ensure the email listing itself comes solely from that skill.
Output Format
- Present an executive summary that highlights the key insights, default and applied windows, most recent action, and top follow-up reminder.
- Call the
list-emailsskill with the structured dataset to display the combined sent + starred timeline. Do not display the email results in any other format. - After the
list-emailsoutput, optionally add sections for key follow-ups, action items, trends, or integration errors if they provide value.
Handling Special Cases
- No sent items: Provide that context in the executive summary and pass only starred entries to
list-emails. - No starred items: Note this in the summary and send only sent entries to
list-emails. - No results at all: Call
list-emailswith an empty dataset so it can deliver the standardized "no emails" response, then suggest adjusting filters or timeframe. - Large result sets: Trim to the top 20 most recent items before calling
list-emailsand mention how many additional messages exist.
Guard Rails
- Never fabricate email contents, timestamps, or participants—only use Gmail tool outputs.
- Do not modify labels or star status; report read-only insights.
- Keep summaries discreet—omit sensitive details unless necessary for context.
- Make all timestamps explicit and timezone-aware.
- If integrations fail, clearly state the error and prompt the user to retry or reauthenticate.
Related Skills
recent-emailsfor broader timeline coverage.starred-emailfor a dedicated starred-only view.
GitHub Repository
Related Skills
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
evaluating-llms-harness
TestingThis Claude Skill runs the lm-evaluation-harness to benchmark LLMs across 60+ standardized academic tasks like MMLU and GSM8K. It's designed for developers to compare model quality, track training progress, or report academic results. The tool supports various backends including HuggingFace and vLLM models.
llamaguard
OtherLlamaGuard is Meta's 7-8B parameter model for moderating LLM inputs and outputs across six safety categories like violence and hate speech. It offers 94-95% accuracy and can be deployed using vLLM, Hugging Face, or Amazon SageMaker. Use this skill to easily integrate content filtering and safety guardrails into your AI applications.
langchain
MetaLangChain is a framework for building LLM applications using agents, chains, and RAG pipelines. It supports multiple LLM providers, offers 500+ integrations, and includes features like tool calling and memory management. Use it for rapid prototyping and deploying production systems like chatbots, autonomous agents, and question-answering services.
